Fraud resistant biometric financial transaction system and method

Abstract

A method and system for authenticating financial transactions is disclosed wherein biometric data is acquired from a person and the probability of liveness of the person and probability of a match between the person or token and known biometric or token information are calculated, preferably according to a formula D=P(p)*(K+P(m)) , wherein K is a number between 0.1 and 100, and authenticating if the value of D exceeds a predetermined value.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS[0001]This application is a continuation of, and claims priority to U.S. application Ser. No. 12 / 444,018, filed on Apr. 2, 2009, which is a National Stage Entry of International Application No. PCT / US07 / 80135, filed Oct. 2, 2007, which claims priority to U.S. Provisional Application No. 60 / 827,738, filed Oct. 2, 2006, all of which are hereby incorporated by reference for all purposes.BACKGROUND OF THE DISCLOSURE[0002]This invention relates to biometric identification and authentication systems and methods, more particularly to authentication for financial transactions using biometrics.[0003]Biometric identification and authentication systems are known in the art, for example systems to compare facial features, iris imagery, fingerprints, finger vein images, and palm vein images have been used. Such systems are known to be useful for either comparing biometric data acquired from an individual to stored sets of biometric data of known “enrolled” ...

AI Technical Summary

Benefits of technology

[0024]A still further aspect and an advantage of the invention is that if a person fails or passes authentication, the person is not informed as to whether non-authentication or authentication was based on probability of liveliness or probability of matching of biometric image. This makes it much more difficult for an attempted fraudster to refine their fraudulent methods since they are not being provided clear feedback.

[0025]As compared to conventional biometric systems and methods, the invention does not merely depend on the probability that the person is who they said they are when authorizing a transaction. The invention includes calculating a second probability which is the probability that the biometric data is from a real person in the first place. The first probability is determined using any biometric algorithm. The second probability is determined using other algorithms which determine whether the biometric data or the person from whom the data is collected is a real person. The decision to authorize a transaction is now a function of both these probabilities. Often, if the first probability is high (a good match), then the second probability typically will also be high (a real person). However, in some cases where a good customer is trying to perform a transaction and the biometric algorithm is having difficulty performing a match (because light is limited for example and the person's web-cam has a low-contrast image), then the first probability could be low but the second probability could still be high.

[0026]The algorithms to determine the second probability (confidence in whether a person is real or not) can be designed to be in many cases less sensitive to conditions out of the control of the algorithms, such as illumination changes and orientation of the person, compared to algorithms that compute the first probability (confidence that the person is a particular person) which are often very sensitive to illumination changes and orientation of the person. Because of this, and since we combine the 2 probabilities to make a decision in a transaction, the reject rate of true authentics can be designed to be greatly reduced.

[0027]The invention authorizes transactions based on a combination of the two probabilities, an attempted fraudster is never sure whether a transaction was authorized or not authorized because they were matched or not matched, or because they were or were not detected as a real person and eliminates the clear feedback that criminals are provided today that they use to develop new methods to defeat systems. As a bi-product, the invention provides an enormous deterrent to criminals since the system is acquiring biometric data that they have no idea can or cannot be used successfully as evidence against them. Even if there is a small probability that evidence can be used against them is sufficient for many criminals to not perform fraud, in consideration of the consequences of the charges and the damming evidence of biometric data (such as a picture of a face tied to a transaction). An analogy to this latter point is CCTV cameras in a high street, which typically reduces crime substantially since people are aware that there is a possibility they will be caught on camera.

[0028]A preferred formula used in calculating a decision whether to authenticate a transaction is D=P(p)*(1+P(m)), where D is the decision probability, P(m) is the probability of a match with a range of 0 to 1, and P(p) is the probability the person is real and the biometric data is valid from 0 to 1. If the algorithm detects person is not live, and no match detected: D=0*(1+0)=0. If the algorithm detects strongly that the person is live, and yet no match is detected: D=1*(1+0)=1. If the algorithm detects strongly that the person is live, and a very good match is detected: D=1*(1+1)=2. If the algorithm detects strongly that the person is live (or more specifically, that biometric data has been collected that can be used by a manual or automatic method after-the-fact to identify the person in prosecution for example), and a poor match is detected of 0.3: D=1*(1+0.3)=1.3 If the threshold is set at, for example, 1.2 for D, then essentially in the latter case, the transaction will be authorized even though the biometric match is not high. This is because the system determined that the biometric data collected can be used by a manual or automatic method after-the-fact to identify the person in prosecution for example. A higher transaction may be authorized if the value of D is higher. Many other functions of Pp and Pm can be used. We use the parallel result to authorize a transaction or access control or other permission, where rejection of a true customer has significant penalty such as a loss of a customer. In the prior art, false rejects and true accepts are often addressed only in consideration of the biometric match performance, and the substantial business consequences of a false reject is often not considered, and therefore few systems have been implemented practically.

[0029]A special advantage of this method and system is that by combining in one algorithm the live-person result with the match result, a fraudulent user does not know whether he or she was authorized or declined as a result of a bad or good match, or because the system has captured excellent live-person data that can be used for prosecution or at least embarrassing public disclosure. The system results in a large deterrent since in the process of trying to defeat a system, the fraudulent user will have to present some live-person data to the system and they will not know how much or how little live-person data is required to incriminate themselves. The fraudulent user is also not able to determine precisely how well their fraudulent methods are working, which takes away the single most important tool of a fraudster, i.e., feedback on how well their methods are working. At best, they get feedback on the combination of live-person results and match results, but not on either individually. For example, a transaction may be authorized because the probability of a live-person is very high, even if the match probability is low. The invention collects a set of live-person data that can be used to compile a database or watch list of people who attempt to perform fraudulent transactions, and this can be used to recognize fraudsters at other transactions such as check-cashing for example by using a camera and another face recognition system. The system also ensures that some live-person data is captured, then it provides a means to perform customer redress (for example, if a customer complains then the system can show the customer a picture of them performing a transaction, or a bank agent can manually look at the picture of the user performing the transaction and compare it with a record of the user on file).

Problems solved by technology

For example, the face biometric is easy to acquire (a web camera for example) but it's ability to tell an impostor from an authentic person is somewhat limiting.

Even though some biometrics such as the iris are sufficiently accurate to have no cross-over between the authentics and impostor distributions when the iris image quality is good, if the iris image is poor then there will be a cross-over and the problem reoccurs.

However, in the field of authentication of financial transactions, high levels of accuracy and speed are critical.

In this field, even a small percentage of rejections of authentics can result in an enormous number of unhappy customers, simply because of the huge number of transactions.

In addition, informing the customer (or attempted fraudster) that they successfully got through a biometric system (or not) is not desirable because it enables fraudsters to obtain feedback on methods for trying to defeat the system.

One problem faced by biometric recognition systems involves the possibility of spoofing.

The prior systems and methods have not achieved significant commercial success in the field of authenticating financial transactions due, in part, from the insufficient speed and accuracy from which prior biometric authentication systems for financial transactions suffered.

More specifically, the current methods of basing a decision to perform a financial transaction on the measure of match means that many valid customers are rejected, due to the finite false reject rate.

This makes it much more difficult for an attempted fraudster to refine their fraudulent methods since they are not being provided clear feedback.

However, in some cases where a good customer is trying to perform a transaction and the biometric algorithm is having difficulty performing a match (because light is limited for example and the person's web-cam has a low-contrast image), then the first probability could be low but the second probability could still be high.

In the prior art, false rejects and true accepts are often addressed only in consideration of the biometric match performance, and the substantial business consequences of a false reject is often not considered, and therefore few systems have been implemented practically.

The system results in a large deterrent since in the process of trying to defeat a system, the fraudulent user will have to present some live-person data to the system and they will not know how much or how little live-person data is required to incriminate themselves.

The fraudulent user is also not able to determine precisely how well their fraudulent methods are working, which takes away the single most important tool of a fraudster, i.e., feedback on how well their methods are working.

Images